The Role of the Tropical Indian Ocean and

Polar Stratosphere for the Sharp Decline of

Antarctic Sea-Ice in 2016

Guomin Wang1, Harry H Hendon

1,Julie M Arblaster

2,3,Eun-Pa Lim

1,S. Abhik

2

1 Bureau of Meteorology, Melbourne, VIC, Australia

2Monash University, Clayton, VIC, Australia

3National Center for Atmospheric Research, Boulder, Colorado, USA

Wang, G., Hendon, H.H., Arblaster, J.M., Lim, E.P., Abhik, S. and

van Rensch, P., 2019. Compounding tropical and stratospheric

forcing of the record low Antarctic sea-ice in 2016. Nature

communications, 10(1), p.13.

Lim, E.P. and Hendon, H.H., 2017. Causes and

predictability of the negative Indian Ocean Dipole and

its impact on La Niña during 2016. Scientific

reports, 7(1), p.12619.

A climate change conundrum:

Artic sea ice has been steadily decreasing but Antarctic sea ice

had been slightly increasing up through 2016

Sep

Oct

2016 Nov

Dec

• Long-term positive trend

• Sharp decline during

September-December 2016

Monthly Antarctic Sea-Ice Extent 1979-2016

IndianO WPO RossS BAS WS

What promoted these atmospheric circulation anomalies?

Anomalous atmospheric circulation in the Antarctic region has

been implicated for the sharp decline (Turner et al. 2017; Stuecker et al 2017)

Circulation and Sea-Ice Concentration

Sep-Oct 2016 Nov-Dec 2016

IndianO WPO RossS BAS WS

High latitude wave 3 pattern Annular (low SAM) pattern

Ozone in the Atmosphere

oC Wm-2

Tropical conditions Sep-Oct 2016

Record negative IOD

Record

strong

convective

dipole

SST OLR

Nino34 DMI OLR DMI

Weak La Nina1979 2016

Ozone in the Atmosphere

Rossby Wave Source and Wave Activity Flux

Sep-Oct 2016

Convective anomaly in Indian Ocean and far western Pacific

appears to be source Rossby wave train

Shading = OLR

Vectors = divergent wind at 200 hPa

Pink contours = Rossby wave source

due to advection of mean absolute

vorticity by anomalous divergent flow

=Vd' dot grad (mean abs vort)

Shading = Streamfunction

Vectors = Rossby wave activity flux

Parallel to Rossby wave group velocity

Divergence: wave source

Convergence: wave sink

Ozone in the Atmosphere

Multiple Linear regression 1979-2015 using DMI and Nino34 as predictors.

Observed behaviour in Sep-Oct 2016 consistent with historical

relationship with IOD/ENSO 1979-2015

OLR, RWS, Div wind Psi200, Wave activity flux

DMI

Nino34

Ozone in the Atmosphere

Display scaled by

DMI 2016

Display scaled by

Nino34 2016

Observed behaviour in Sep-Oct 2016 consistent with historical

relationship with IOD/ENSO 1979-2015

OLR, RWS, Div wind Psi200, Wave activity flux

Neg IOD

was

primary

source of

wave 3

pattern,

boosted

by La Nina

in SA

sector

Obs 2016

DMI

Nino34

a

Ks=sqrt(B*/ubar)

Wave 3 pattern understood from Rossby wave theory

(e.g. Hoskins and Ambrizzi)

Wave train refracted into high latitude wave guide

U200

Beta*

Ozone in the Atmosphere

Easterly wind

anomalies (low

SAM)

>Southward

(warm) Ekman

transport

Sea ice

concentration

anomalies

High SAM is usually observed during La Niña. Why did a

strong low SAM rapidly develop during Nov-Sep 2016?

November-December 2016 Circulation: Low SAM rapidly

developed and continued to promote sea ice decline

Ozone in the Atmosphere

Impact of MJO event during November 2016

4 NOV

MJO acted to shut down the convection in east IO in early Nov

that was providing the Rossby wave source during Sep-Oct.

Ozone in the Atmosphere

Daily SAM index

Polar Cap Zonal mean

zonal winds 50-70S

Also experienced a strong early breakdown (weakening) of the

polar stratospheric vortex that first emerged in upper

stratosphere in Oct that then coupled downward through Dec

Observed 2016

Historical relationship 1979-2015 : regression

of polar cap zonal winds onto Nov-Dec

surface zonal wind: downward coupling is a

prominent feature of circumpolar winds in late

spring-summer

Dot proportional to

sea ice drop (red)

or increase (green)

OLR Dipole Sep-Oct

Circumpolar

westerlies

Sea ice decline

2016 was result

of

unprecedented

negative IOD

followed record

negative SAM:

Just bad luck?

Partial recovery in 2017 suggests that internally generated

variability (IOD and polar vortex weakening) played a primary role

in the 2016 decline: but doesn't preclude a role for global warming

Causes and predictability of the record strong negative Indian Ocean dipole 2016

Eun-Pa Lim & Harry H. Hendon

Science to Services

Bureau of Meteorology

Was there a role for ongoing warming in the Indian Ocean?

Background

Strong La Nina was expected in mid-2016

- Strong discharge of the heat in the tropical Pacific subsurface in the 1st quarter of 2016 from massive El Nino 15/16

- Early development of strong –ve IOD by July

But La Nina of 2016 ended up to be a weak event

Aim of the study:

Causes and predictability of record strong negative IOD 2016

(taken from BoM seasonal outlook meeting materials; E. Miles)

WWV

nin

o3

4

Negative IOD in June – Sep 2016

Typically IOD peaks in Sep-Nov

Negative IOD of 2016 developed from May, peaked in July and Sep & decayed rapidly in November

Strongest negative IOD June-Sep since 1960-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

1 2 3 4 5 6 7 8 9 10 11 12

An

om

aly

(deg

C)

Calendar Months

DMI2016STDD…

Preconditions of negative IOD in June-Sep 2016Record high SSTs in the tropical IO in March-May 2016

Basin-wide warming and downwelling Rossby wave forced by strong El Nino 2015-16

But also significant contribution from long-term warming trend

Late boreal spring IO temperature trend pattern is similar to negative IOD pattern

Temperature trend over 1960-2014 in April/May

Conduct coupled model seasonal prediction

experiments to elucidate the role of warming

trend and other mechanisms for promoting the

record strong negative IOD

Coupled model seasonal forecast sensitivity experiments

POAMA: an atmosphere-ocean coupled system, run operationally in BoM for sub-seasonal to seasonal climate outlook:

• Atmospheric model: Bureau’s Atmospheric Model v3 (~250km x 250km x 17 vertical levels)

• Ocean model: Australian Community Ocean Model 2 (200km x 50-150 km x 25 vertical levels)

Initialisation:

Observed atmosphere, land and ocean initial conditions generated from BoM's data assimilation systems - ALI (Hudson et al. 2010 Clim. Dyn.) and PEODAS (Yin et al. 2011 Mon. Wea. Rev.)

Skill has been assessed using 33 member ensemble hindcasts for 1980-2014

POAMA is competitive with other international models in predicting ENSO and IOD (e.g. Barnston et al. 2012, Shi et al. 2012) gives us confidence to use POAMA to understand the dynamics of the IOD and La Nina of 2016

POAMA forecast sensitivity experiments

7 experiments: a CTRL forecast, 5 forecast sensitivity exps, and a climatological exp

Each experiment consists of 11 member ensemble forecasts

In all experiments, forecasts were initialised on 21st April 2016 and verified for May to November 2016

In the 7 experiments, atmosphere & land were initialised with observed conditions of 21st April 2016

Ocean was initialised with 7 different configurations (various combinations of observed conditions of 21st April of 2016 plus 21st April climatology 1980-2010

Design of experiments – different ocean initial conditions

CTRL: Observed conditions of 21 Apr 2016 (i.e. Bureau's real-time fcsts)

DTRND Exp: Observed conditions but temperature and salinity trends over 1960-2014 removed

Trend on 21 Apr 1960-2014

SST trend: significant warming over most of the Indian and the western Pacific sea surface

Eq. subsurface temperature trend:

* deepening of the thermocline in the eastern IO

* shallowing of the thermocline in the eastern Pacific and eastern Atlantic subsurface

Design of experiments – different ocean initial conditions

IO Exp: observed conditions in the Indian Ocean & climatological conditions elsewhere

IOAO Exp: observed conditions in the Indian Ocean and Atlantic Ocean

IOPO Exp: observed conditions in the Indian Ocean and Pacific Ocean

ClimIO Exp: climatological conditions in the Indian Ocean & observed conditions elsewhere

Negative IOD of 2016

• POAMA CTRL exp skilfully predicts the –ve IOD of 2016

• The –ve IOD-like long-term ocean T trend +velycontributed to the strength of the 2016 –ve IOD

• Using observed ocean initial conditions only over the Indian Ocean was good enough to generate strong –ve IOD

• Adding observed Atlantic Ocean or Pacific Oceaninformation didn't make any difference

• Climo Indian Ocean initial conditions, -ve IOD was not predicted at all

Strong Jun-Sep –ve IOD of 2016 was primarily driven by the Indian Ocean conditions with moderate contribution from trend

ctrl

dtrend

obs

Causes of the negative IOD 2016

2016

Composite of five strongest Jun to Sep –ve IOD1992,1996, 1998, 2005, 2010

No signature of preceding downwelling K-wave and much weaker westerly wind burst in May

Outstanding features of –ve IOD 2016

Ocean subsurface wave dynamics

: Eq downwelling Kelvin wave emanating from west boundary in Feb resulting from 2015-16 El Nino

Air-sea feedback

initiated with westerly wind bursts over central IO in May

20C Isthm Depth SST U10

Lim & Hendon (2017) Sci. Rep, 7, 12619

Summary of IOD Experiments

Strong negative IOD was skilfully predicted by POAMA CTRL experiment initialised in late April 2016; predictability provided by antecedent conditions in Indian Ocean

Downwelling oceanic Kelvin wave (leftover from 2015-16 El Nino)

Negative IOD-like long-term temperature trend contributed to the extraordinary strength of this negative IOD

• The sharp decline in Antarctic sea-ice extent in Sep-Oct 2016 promoted

by record negative Indian Ocean Dipole (IOD) event

• Emphasizes the important role that tropical Indian Ocean plays for

global climate

• Random occurrence of polar stratospheric warming in late Oct then

maintained the ice decline in Nov-Dec by promoting low SAM

• General conclusion is that dramatic sea ice decline was a result of

internal ocean-atmosphere variability

• However a possible role of climate change is suggested via promotion of

the strong negative IOD by ongoing warming of the Indian Ocean

• Our results suggest that strong negative IOD events maybe more

likely in the future.

Overall Summary

2016

Composite of four strongest La Nina1988, 1998, 2007, 2010

Causes of weak La Nina of 2016

20C Isthm Depth SST U10

Outstanding features of La Nina 2016

- Extraordinary long-tail of El Nino especially over the dateline

- Much delayed air-sea coupling for La Nina

- Forecasts initialised with the warming over the dateline produced weaker La Nina than those without the warming

The cold condition over the NINO34 region was weak in the 2nd half of 2016

• In CTRL exp, POAMA over-predicted La Nina development initially but predicted it better from August onwards

• Cooling trend in the eastern Pacific subsurface caused a weaker La Nina, but the difference between CTRL and DTRND is not statistically significant (< 90%c.l.)

• La Nina was better predicted with observed initial conditions used only over the Indian Ocean

• Adding realistic Pacific Ocean or Atlantic Ocean information weakened the strength of La Nina forecast

• Without realistic Indian Ocean initial conditions, La Nina was not predicted

Indian Ocean played a key role in driving this La Nina of 2016

La Nina of 2016